What Do Retweets Indicate? Results from User Survey and Meta-Review of Research

What do Retweets indicate? Results from User Survey and Meta-Review of Research

P.T. Metaxas, E. Mustafaraj, K. Wong, L. Zeng, M. O’Keefe, S. Finn Computer Science, Wellesley College {pmetaxas,emustafa,kwong,lzeng,mokeefe,sﬁnn}@wellesley.edu

Abstract to ask whether it is interesting to study any specific social media platform. If Twitter does not exist in a few years, will Arguably one of the most important features of Twitter is the support for “retweets” or messages re-posted verbatim by a this paper’s findings matter? We argue that the answer is yes: user that were originated by someone else. Despite the fact Twitter is recording human communication that requires rel- that retweets are routinely studied and reported, many impor- atively little effort to produce and consume. This communi- tant questions remain about user motivation for their use and cation is what we want to study. While any particular social their significance. In this paper we answer the question of media platform may cease to exist or lose popularity in the what users indicate when they retweet. We do so in a compre- future, the importance of human interaction through social hensive fashion, by employing a user survey, a study of user media is unlikely to change. Humans are social animals and profiles, and a meta-review of over 100 research publications their desire to communicate with each other and comment from three related major conferences. Our findings indicate on their social environments is one of their universal and that retweeting indicates not only interest in a message, but unique characteristics. We have chosen to study the interac- also trust in the message and the originator, and agreement with the message contents. However, the findings are signif- tion of humans through social media in an abstract way, not icantly weaker for journalists, some of whom beg to differ a way specific to the particular social media platform. We declaring so in their own user profiles. On the other hand, the are simply looking at the behavior as revealed through their inclusion of hashtags strengthens the signal of agreement, es- interactions. Further, its service makes it easy for people to pecially when the hashtags are related to politics. While in say something. The effort in contributing to the general so- the past there have been additional claims in the literature cial dialog is far less than that of writing a comment a blog about possible reasons for retweeting, many of them are not on a web site or a newspaper op-ed, and it has wider im- supported, especially given the technical changes introduced pact than talking person-to-person or via email. In addition, recently by Twitter. the effort to propagate a message sent by someone else is also remarkably small – giving rise to degrading character- 1 Introduction izations of online social participation such as “slactivism” Twitter is a real-time information network that allows its (Gladwell October 4 2010). users to write short messages (“tweets”) up to 140 char- So, why do people retweet? The published literature does acters in length. We chose to focus on Twitter in particu- not give a conclusive answer to this question, though it was lar because, unlike other popular social media platforms, it studied as early as 2010 (Boyd, Golder, and Lotan 2010). has grown into a real-time news source created by everyday Some of the claims made in the past can be characterized users. Twitter is credited for its role in political events such as straightforward self-evident, (e.g. that people retweet to as monitoring elections, aiding the so-called “Arab Spring” broadcast information) and others as outdated (e.g. technical (Bruns, Highfield, and Burgess 2013), and for drawing at- changes in Twitter do not support them anymore, ie. that tention to news stories that were initially largely ignored by people retweet to appropriate information). However there is traditional news media such as the Wendy Davis filibuster1 no agreement in the literature whether retweeting indicates and the Michael Brown shooting in Ferguson, MO. trust, agreement or even endorsement of the originator or the Over the years Twitter has developed its own syntactic message. Moreover, there is no clear understanding how the components. This paper focuses primarily on retweeting: retweeting rates are affected by emotion, political intention, Users retweet (RT) to forward a message from another or propaganda, if they are affected at all. Our paper tries to source to their own followers. answer as conclusively as possible these questions. As a platform, Twitter has several characteristics that The remainder of this extended abstract is organized as make it convenient for research. Yet, it is a valid question follows: In section 2 we present the results of a user sur- Copyright c 2015, Association for the Advancement of Artificial vey that answers the question of what Twitter users think Intelligence (www.aaai.org). All rights reserved. about their practice of retweeting. In subsection 2.2 we study 1Texas filibuster on abortion bill rivets online, by Heather Kelly. an important user subgroup, reporters, that includes mem- CNN.com, June 26, 2013. bers whose retweeting practice appears to differ significantly from that of the general user. We also present an analysis 2.1 Survey Results of user profiles that include a disclaimer about retweeting We launched the study by utilizing our social links and not indicating endorsement. Section 3, truncated in this ex- reached directly audiences, primarily in academic and ethnic tended abstract, presents a meta-review of research papers communities that we had social and professional links (Twit- from three major conferences. Finally, section 4 has our con- ter, Facebook, Linkedin and email). Reflecting on the mode clusions. of launching our survey, the responders’ sample (n=316) was highly educated, with 25.3% possessing a Bachelor’s 2 User Survey degree and an additional 43% a Master’s Degree or higher. Of the responders, thirty-three (11.4%) identified them- In mid-July 2014, we conducted a survey asking Twitter selves as bloggers, journalists, or reporters. users about various aspects of their retweeting behavior. The Our sample was representative of both males (43%) and survey responses indicated that there is general consensus on females (54.4%), with 2.6% choosing not to disclose gen- several important factors that the majority of Twitter users der. In terms of ethnicity, we left it as an open-ended ques- consider when deciding whether or not to retweet a given tion and recorded participants responses as Caucasian/White message. However, we also found that there were also sig- (46%), Asian or Pacific Islander (11%), Black (3.8%), other nificant variations across subcategories of users, including (22%), and non-reporting (17%). differences across the responses from reporters and non- The age demographics of our data is consistent with Pew’s reporters, individuals interested in politics, users of differ- report of Twitter users2 on several respects. Reflecting the ent ages, and users who use Twitter with different levels of similar trend that younger individuals tend to be on social frequency. media platforms more than older ones, including Twitter, We designed a comprehensive twenty-one item question- 38.6% of the participants report being under the age of 24, naire (http://bit.ly/TwitterUsageSurvey) that 25.9% between 25-34, 13.9% between 35-44, 12.3% be- asked participants about their retweeting behavior. Initial tween 45-54, and 9.2% over the age of 54. Additionally, our survey questions inquired about whether the participant had data captured users who use Twitter on average as 42.2% of a Twitter account, and if so, for how long. Other questions participants use Twitter daily, 23.8% use it weekly, and 34% relating to general Twitter usage included how often the par- use it monthly or less. 55.7% have less than one hundred ticipant uses Twitter, how often they tweet, why they use followers while 8% have over one thousand. 40% of partic- Twitter, how many followers they have, and how many ac- ipants provided their Twitter handle. Figure 1 summarizes counts they follow. The survey also asked participants about some of our results about participants primary motivations what factors they find important when deciding whether or for using Twitter and their retweeting behavior. not to follow an account on a scale from 1 (not at all) to 5 (extremely). After answering these questions, we asked participants to indicate how often they retweet, if at all. Only participants who responded that they are familiar with retweeting were led to a page of questions asking them about what factors, both about the message and the account, they consider important when choosing to retweet a post. In terms of the message, options were whether they find the message interesting, emotionally resonant, entertaining, something they endorse, something that they agree with, something their followers might find interesting, something formative or something they find trustworthy. In terms of the account, options were whether they find the account credible, a celebrity, someone they know personally, someone who shares their opinion, someone they support, or someone they trust. Both cases were multiple-choice ending with a write-in (“other”) option. These questions were posed as statements on a likert- scale from 1 (strongly disagree) to 5 (strongly agree). They were also asked about what is more important when they decide to retweet: the message, the account or both. Additionally, participants who had reported retweeting in Figure 1: Summary of the results of our survey why people the past week were asked to recall certain features about use Twitter (top), and how often do they retweet (bottom). one particular retweet of their choice. These additional questions were designed to demonstrate whether users’ actual be- When it came to retweeting behavior, the majority of par- havior is consistent with their overall impressions of their ticipants responded that they believed that the message was own behavior. The final questions captured basic demographic questions about the participants including age, self- 2Pew Research Internet Project: Social Networking Fact Sheet. described ethnicity, gender and educational level. Data from 2014. more important than the account when choosing to retweet whether they agree with the given message as other factors a post. That is, when choosing whether or not to retweet a that were important (see Fig. 2, top). tweet, they evaluated and put more significance on the tex- Some of the initial insights that we found helped guide our tual content of the tweet than on the user who had written further exploration of the data that was collected. We found the tweeted. In fact, some users even reported to not pay at- that the majority of participants tended to retweet posts from tention to who tweeted the post in the first place. accounts that they follow (vs, accounts that they do not follow) and therefore it is important to understand why they choose to follow certain accounts in the first place. Consistent with the hypothesis that users care about credibility and personal interest when navigating through social media platforms for information, responses to the survey indicated that specific factors including “shares my interest” and “trustworthiness” were most commonly cited as important factors to consider when choosing whether or not to retweet a given account (see Fig. 2, bottom). Meanwhile, factors such as “celebrity” and “entertainment” were not rated as important in Twitter users’ decision-making pro- cess. Additionally, there were other factors that yielded responses similar to that of “shares my opinion” where there was no strong opinion. The participants’ responses were recorded from a likert-scale from 1 (strongly disagree) to 5 (strongly agree) to “agree,” “neither”, and “disagree.” Reflecting the amount of emphasis the participants reported placing on the account as opposed to the message, results regarding what qualities about an account are important to consider when choosing to retweet a message indicated less agreement and certainty. The criterion that seemed most important overall was that the account is “credible.” Our survey did not ask about how they judge whether or not an account is credible and whether or not that is truly the case. However, this suggests that Twitter users do care about whether or not a given account appears trustworthy.

On the disclaimer “Retweeting does not mean endorsement” In light of the previous findings, of particular interest is the statement encountered in certain Twitter users’ Figure 2: (top) 94% of survey participants indicate that per- profile summary that retweeting does not mean endorse- sonal interest in a message is the primary reason for retweet- ment or agreement. It seems that this statement contradicts ing. Reasons that get very strong support (about 3/4 of the findings presented in this section, so we examined it in users) are trustworthiness, informativeness and user agree- greater depth. There are relatively few users who include ment with a message. More than 2/3 of users also indicate such statement. Using the Twitter Search interface for the as reasons to retweet curation (interest to their followers), terms “RT, endorsement” we found 384 accounts containing message endorsement and entertaining nature of the mes- both of these terms. An exhaustive categorization of these sage. Emotional resonance with the message was not viewed account showed that 197 (52.7%) of them belong to peo- as an important reason. (bottom) Participants indicated that ple in the Media (reporters, journalists, media producers) in- account credibility (76%) and trustworthiness (59%) are the cluding 11 bloggers, 99 (26.5%) to politicians, 43 (11.5%) most important factors about the account when deciding to to non-media related companies, 24 (6.4%) to people who retweet a post. Sharing opinion with and support of the orig- list politics in their interests, and 9 (2.4%) to academicians. inator were not viewed as important reasons, especially for The remaining 0.5% of accounts were from people stating reporters. Little or no support has personal knowledge of the that their RT is, in fact, endorsement. originator and whether the originator was a celebrity. We then tried to collect a comprehensive sample of such profile statements. Using the Twitter API we collected all Participants provided consistent responses when respond- the user profiles that make a statement along these lines. In ing to survey questions that asked about various factors they fact we found that the terms “retweet”, “RT” co-occur with considered important about a given message when choos- the terms “agreement” or “endorsement” in 2,585 profiles. ing whether or not to retweet it. The majority of partici- We randomly sampled uniformly 10% of them and found pants indicated that factors such as how interesting the mes- that media-related accounts comprised 52.65% and 12.65% sage is to themselves as well as their followers are impor- to politicians. There is an over-representation of journalists, tant. They also cited trustworthiness, informativeness, and reporters and media producers who include this disclaimer in their profiles. Other journalists have challenged this prac- test for the difference in the amount by which political users tice3 since its inclusion in one’s profile is not an effective versus non-political users on Twitter tend to retweet ac- disclaimer (many people will likely miss it). counts that they don’t follow was not significant. Finally, one has to consider the reason for people placing such a clarification statement in their profile as addressing 3 A Meta-Review of research papers a concern that one’s retweeting practices may be mistaken To focus our research question a bit further, we did a meta- by others. It is reasonable to argue that the mere fact of review of the research corpus that has been published in including this disclaimer is an implicit admission by those the last several years on Twitter. Starting with abstract in- using it that others may mistake their intentions, because spection, we collected over 100 relevant papers published for most people retweeting is endorsement. Journalists and from 2008 to 2013 in three major conference venues re- politicians may worry about the effect of their retweets on lated to the subject: AAAI ICWSM, IEEE SocialCom, their curated public images. and WWW. From the initial core we expanded the cov- erage to references found in the papers from these three 2.2 Reporters and Users interested in Politics venues. To obey the page limit, we refer the interested reader As we mentioned, several journalists on Twitter include in to http://bit.ly/RT-meaning-metareview. We their Twitter profile description a disclaimer stating that simply state that our findings indicate support in the survey “retweeting does not equal endorsement.” It’s almost as if results, and provide insight on how the emotions associated this subgroup of individuals feel a need to distinguish them- with hashtags can affect the retweeting rates and visibility of selves from the attitudes and behavior of other general users a topic being discussed on Twitter. on Twitter. To see if participants in the study responded in ways that reflected this distinction, we ran five Welch Two 4 Conclusions Sample t-tests, as well as the Wilcoxon rank sum tests, to While in the past there have been several claims in the confirm the t-tests. While the t-test’s assumption of normal- literature about possible reasons for retweeting, there was ity is fairly robust, we wanted to confirm that we would gen- no focused study on the issue. This paper aims to close erate similar results even if that assumption is not made. this gap. Employing a user survey, a study of user profiles, There was a significant difference in the degree to which and a meta-review of over 100 research publications from reporters (M= 3.4, SD=1.1) versus non-reporters (M= 4.1, three relevant major conferences, we answer the question SD=0.9) agreed with the statement that “When retweet- of what users mean when they retweet. Our findings show ing, it is important that the message is something that I that retweeting indicates, not only interest in a message, but agree with,” t(32.3)=-3, p=0.006. We also found that 57% also trust in the message and the originator, and agreement of reporters agreed and 17.9% disagreed with the statement, with the message contents. However, the findings are sig- compared to 75% of non-reporters agreeing and 3.8% dis- nificantly different for journalists, who are more likely to agreeing with the statement. include a disclaimer on their user profiles that their retweet- There was also a significant difference in the degree ing does not mean agreement or endorsement. The inclusion to which reporters (M= 3.3, SD=1.1) versus non-reporters of hashtags strengthens the signal of agreement, especially (M=3.9, SD=0.9) agreed with the statement that “When when the hashtags are related to politics. retweeting, it is important that the message is something that While this study presents an extensive sample size, we ac- I endorse,” t(32.7)=-2.9, p=0.007. While 57.1% of reporters knowledge that there are limitations to this research. For ex- agreed and 21.4% disagreed with the statement, 69.5% of ample, there are under-representations of some demographic non-reporters agreed with the statement and 6.6% disagreed. groups and focus on a relatively highly-educated sample Finally, there was some difference in the degree to which which contribute to a potential limited generalizability of the reporters (M= 3.7, SD=1.0) versus non-reporters (M= 4.0, study. Additionally, there is a possibility that there is some SD=0.9) agreed with the statement that “When retweeting, social desirability bias due to the nature of self-reporting. it is important that the account is one that I find credible,” However, we do not believe that either of these limitations t(35)=-1.88, p=0.07. While 72.4% of reporters agreed and have substantial impact on our findings. 13.8% disagreed with the statement, 76.2% of non-reporters Acknowledgements.This research was partially sup- agreed with the statement and 5.7% disagreed. ported by NSF grant CNS-1117693 and by the Wellesley There was a significant difference in the degree to which Science Trustees Fund. political users (M= 3.6, SD=1.1) versus non-political users (M= 3.1, SD=1.1) on Twitter find it important to follow ac- References counts that share their opinion, t(116)=2.7, p=0.008. While Boyd, D.; Golder, S.; and Lotan, G. 2010. Tweet, tweet, 76.1% of political users were motivated and 4.5% believed retweet: Conversational aspects of retweeting on twitter. In that when choosing to follow certain accounts it was impor- HICSS, 1–10. IEEE. tant that the accounts share their opinions, 71.6% of non- Bruns, A.; Highfield, T.; and Burgess, J. 2013. The political users believed it important and 5.9% did not. The arab spring and social media audiences. Am Behav Sci 57(7):871–898. 3E.g., see Commentaries at the Washington Post http://wapo.st/19OIcfy and NPR’s “Retweets are endorsements” Gladwell, M. October 4, 2010. Small change: Why the http://bit.ly/YJBQgQ and http://bit.ly/1v9lBY3 revolution will not be tweeted. The New Yorker.